Pedestrian and cyclist safety in the context of micromobility on the shared-use paths

Before the main measurements, several short test recordings were conducted to determine the optimal camera settings, angle, and position. Measurements were scheduled during peak usage periods on shared-use paths, typically from late April to September, in dry, sunny weather between 1 PM and 6 PM. Some locations were near cultural or social events, which increased user activity, though the measurement points were far enough to avoid affecting user trajectories. A video camera was mounted on a high tripod placed at least 2 meters from the path, recording from an elevated side view for around four hours per location. Fixed reference points were marked along the path, spaced 3 meters apart, and the path’s width was measured. These references were used to transform the video into geographic coordinates for analysis. All video recordings were standardized to 1280×720 px at 25 fps. Object detection of pedestrians and cyclists was performed using the YOLO v9 algorithm, and tracking was done using BoT-SORT. Trajectories were generated from bounding boxes and transformed using affine transformation.Data were collected at 10 locations across the Czech Republic and were manually validated. Key metrics such as trajectory length, average speed, movement direction, and time were calculated. Intersections and overlaps of trajectories were analyzed, with time and distance of interaction recorded. Conflicts were defined as observable situations with a potential risk of collision if participants didn’t adjust their movement. All measured characteristics were segmented into 15-minute intervals for further analysis.